Research 2 (Revision)

The Value of Text for Waiting Time Prediction using deep learning: A Repair Shop Case Study 

One of the most formidable tasks for service companies is to inform customers about the approximate waiting time. Overestimated waiting time may force customers not to proceed with a company, and underestimated waiting time may lead to an impossible responsibility for the company to accomplish. In both cases, customer dissatisfaction would occur. To tackle this critical challenge, we propose a two phases predictive model using textual combined with structured data to predict waiting times as a regression task in service systems. In the first phase, the Bag of Words (BoW) as a text mining approach is utilized to transform texts into numbers. In the second phase, based on nearly 31,000 case study data, our work exploits a Multi-Layer Perceptron (MLP) architecture from deep learning as the best predictive model for a regression task. Our numerical results show that deep learning by having 87.61% accuracy has the best performance compared to conventional machine learning algorithms for predicting a continuous output based on textual data. Also, our paper illustrates the value of texts in waiting time prediction since results show that involving textual data can elevate prediction accuracy by almost 5.2%. Finally, we demonstrate that our predictive model can assist experts' decisions in underestimated waiting time cases by improving 50.12% accuracy. Also, from the managerial perspective, we illustrate how textual data are vital in terms of descriptive analysis to find deficiencies in a service system.

Word cloud of descriptions

Accuracy boxplot of all models in 50 iterations Example Text

Predicted waiting time histogram of each model for the test sets Example Text

This research was honored to receive the third-best research paper among 130 candidates at POMS international conference 2022:

Presentation of the paper in POMS International Conference 2022: